A Domain-Specific Architecture for Elementary Function Evaluation

Table 1

Accuracy, throughput, and table size (for SPU/double precision).

Function	Cycles/double; new	Cycles/double; SPU	Speedup (%)	Max error (ulps)	Table size ()	Poly order
recip	3	11.3	376	0.500	2048	3
div	3.5	14.9	425	1.333	recip	3
sqrt	3	15.4	513	0.500	4096	3	18
rsqrt	3	14.6	486	0.503	4096	3
cbrt	8.3	13.3	160	0.500	8192	3	18
rcbrt	10	16.1	161	0.501	rcbrt	3
qdrt	7.5	27.6	368	0.500	8192	3	18
rqdrt	8.3	19.6	229	0.501	rqdrt	3	18
log2	2.5	14.6	584	0.500	4096	3	18
log21p	3.5	n/a	n/a	1.106	log2	3
log	3.5	13.8	394	1.184	log2	3
log1p	4.5	22.5	500	1.726	log2	3
exp2	4.5	13.0	288	1.791	256	4	18
exp2m1	5.5	n/a	n/a	1.29	exp2	4
exp	5.0	14.4	288	1.55	exp2	4
expm1	5.5	19.5	354	1.80	exp2	4
atan2	7.5	23.4	311	0.955	4096	2	18
atan	7.5	18.5	246	0.955	atan2	2 + 3
asin	11	27.2	247	1.706	atan2	2 + 3 + 3
acos	11	27.1	246	0.790	atan2	2 + 3 + 3
sin	11	16.6	150	1.474	128	3 + 3	52
cos	10	15.3	153	1.025	sin	3 + 3
tan	24.5	27.6	113	2.051	sin	3 + 3 + 3